Geologia CroaticaGeologia Croatica Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina  Dušan Polomčić and Dragoljub Bajić Department of Hydrogeology, Faculty of Mining and Geology, University of Belgrade, Belgrade, Djusina 7, Serbia; (osljane@orion.rs) doi:10.4154/gc.2015.07 ABSTRACT This paper presents groundwater management based on the results of groundwater modeling. Model used to simulate the effects of the operation of three dewatering system scenarios of the Buvač open cast mine. Several aquifers exist within the region of the Buvač limonite ore deposit, which puts ore exploitation at risk due to water ingress. Ground- water modeling was used to understand the groundwater regime within these aquifers. A previous numerical model (2007) was completed for the state of the groundwater regime prior to the opening of the open-cast mine. In the last five years, development of the open cast mine and associated activities have drastically modified the flow field. The model was modified for the altered conditions on site, and with new data from monitoring both the ground and sur- face water regimes. The existing system of protection of the cast mine from groundwater encompasses drainage of both the alluvial and ore body aquifers. This paper analyzes three scenarios for protecting the opencast mine from groundwater between 2013 – 2024. The new model was used to simulate the effects of operating the three proposed dewatering systems. Differences in the operating effects of the scenarios were quantified through the conditions of groundwater levels and water budget. Recommendations have also been given for performing prediction calculations in the case of the existence of drainage wells, with the most effective dewatering system determined from a hydro- dynamic aspect. The results presented may be used for developing an effective dewatering system. Keywords: numerical modeling, groundwater management, mine dewatering, Bosnia and Herzegovina 2000; LI et al., 2014). These solutions were often developed based on simplified assumptions and boundary conditions that restrict their applicability in real world mining situa- tions. Numerical modeling codes such as Modflow (HAr- BAuGH et al., 2000) enable finding solutions to significant regional groundwater problems and are very important for the design of dewatering systems for open cast mines. Nu- merical models do not have the limitations of analytical so- lutions and are suitable for the simulation of all aquifer con- ditions (rApANTovA et al., 2007; poloMčić et al., 2013b). For example, BoSkiDiSA et al. (2012) describe the use of Modflow to examine several alternative scenarios for 1. INTRODUCTION Groundwater ingress into a mine during its operative life is one of the most important mining concerns. It is therefore necessary to quantify groundwater inflow into the mine and the effectiveness of the dewatering system in operation. prior to the development of computational numerical mod- eling methods, groundwater impacts were predicted by ana- lytical and empirical methods, or hard-won experience. Many analytical solutions for predicting water inflow into mine excavations can be found in the literature (HANNA et al., 1994; MAriNelli & Niccoli, 2000; SHeveNell, Geologia Croatica 68/2 123–137 11 Figs. 2 Tabs. Zagreb 2015 mailto:osljane@orion.rs Geologia Croatica 68/2Geologia Croatica 124 The study area encompasses the wider region of the Buvač open cast mine. in order to design a dewatering sys- tem, a numerical model of the open cast mine (poloMčić et al., 2013a) was constructed when the mine was opened in 2007, with input data from the period 1971-1973. The use of data available from the beginning of the 1970s decreased the model’s reliability; however, these were the only data available at the time of model construction. From october 2010 to october 2012, new hydrogeological research and groundwater monitoring was conducted. In the meantime, from 2008, ore exploitation commenced at the open cast mine. As one of the measures of protection of the open cast mine from groundwater and surface water, the Rivers Go- mjenica and Bistrica were diverted into the new river bed of the River Gomjenica, to the north of the mine (Figure 1). The first numerical model was corrected and adjusted to the new conditions in the study area. New data regarding the site have enabled certain improvements to be made in assigning boundary conditions in the model. The model was supple- mented with a set of the new registered data on the ground- water and surface water regimes. A new recalibrated model was used to simulate the operational effects of the dewater- ing systems from 2013 - 2024, i.e. until the end of ore ex- ploitation. In total, three scenarios operating different dewa- tering systems were analyzed. The most efficient system, from a hydrodynamic perspective, was chosen by compar- ing the predicted operational effects. groundwater extraction from two aquifers, and describe how the calibrated model allowed for better management of groundwater resources. MylopouloSA et al. (2007) de- scribed an integrated water resources management plan wherein a numerical groundwater model of a deep aquifer system was used to develop alternative exploitation plans for the deeper aquifer. Dewatering systems should also be designed and con- structed to fully protect groundwater and surface water from mining impacts, to enable the sustainable extraction of ore. The new age of numerical analysis has enabled modeling of increasingly complex hydrogeological regimes and inclusion of components such as dewatering systems, drainage wells, and cut-off walls (STrzoDkA & FiScHer, 1988; ŠuBArANović et al., 2013). Numerical models are most often used for open cast mines in order to determine the in- flow of water into the mine workings (BrAwNer, 1982), for assistance in designing dewatering systems (STruziNA et al., 2011; BAHrAMi et al., 2014) or in order to analyze the effects and optimization of operating several possible dewatering systems (wu et al., 2010; JiANG et al., 2013). They are also used to analyze the environmental impact of mining activities (wu et al. 2000, QiAo et al., 2011; MA- rANDi et al., 2013) or to predict groundwater mounding after ore exploitation and dewatering system operation ends (BANkS et al., 2010; GöDeke, 2011). Figure 1: Location and test borehole map of the study area. Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 125 Due to the changed conditions in exploration area, because of the mine dewatering and the beginning of min- ing, recalibrating of the first model was done (poloMčić et al., 2013). Additional advantages of the recalibrated model, compared to the first model, arise from the follow- ing facts: • more recent detailed exploration, which included ex- pansion of the observation network of piezometers, resulted in a better understanding of the groundwater regime • the model was calibrated both manually and automat- ically, using peST (DoHerTy & HuNT, 2010) soft- ware, and the hydraulic properties which constituted model inputs were thus optimized. This contributed to a more efficient and more reliable approach to groundwater modeling and, as a result, recali- bration of the new model was implemented at a higher level than that of the first model (poloMčić et al., 2013a). 2. THE HYDROGEOLOGICAL SETTING OF THE STUDY AREA The Buvač limonite ore deposit (Figure 1) is located in Bos- nia and Herzegovina. it belongs to the northeastern part of Figure 2: a) Hydrogeological cross-section A-A’ (profile line in Fig. 1) and b) 3D aquifer model in the study area. Geologia Croatica 68/2Geologia Croatica 126 the ljubija Metallogenic province and occupies the south- eastern peripheral area of the omarska mineral ore deposits which, in addition to the Buvač mine, comprise the open cast mines of Jezero and Mamuze, (separated by about 500 m). The omarska ore deposits lie in a morphological depression of the omarska-prijedor field, which trends Se-Nw and is about 5 - 10 km wide and bout 20 km long. The slopes of Mt. kozara, which separates the prijedor Gorge from the Sava valley and the lower end of the pannonian Basin, con- stitute the northern boundary of the omarska-prijedor field. Along the southern, southwestern and northern perimeters, the terrain continuously rises gradually to Mt. Grmeč, on the left bank of the Sana river, and Mt. Menjača on the right bank of the Sana river. The terrain is open to the north. The deposit is of carboniferous age and extends over an area of about 3000 m2 with an average thickness of about 20 m. The ore body is oriented northeast-southwest with a dip towards the northeast. It lies very close to the surface on the south- west side, while on the northeast side it is located at a depth of about 170 m (Figure 2a). The lithological characteristics were determined based on core logging in more than 550 boreholes (Figure 1). From a hydrogeological perspective, the basic characteristic of the study area is the variability of hydraulic conductivity with depth. within the study area, there are four types of aquifers: a Quaternary alluvial aqui- fer, another in pliocene sands, an aquifer within the limonite ore body, and a confined karst aquifer located at the bottom of the ore body. The interrelationship between these aquifers is shown in the hydrogeological cross-section A-A’ (Figure 2a) and in the 3D aquifer model (Figure 2b). The alluvial aquifer is formed in alluvial sediments (Quaternary gravels) of the River Gomjenica and its tribu- taries. This aquifer is unconfined, and there is good evidence that it is connected hydraulically with the river, as shown in fluctuations in the Gomjenica stages and the water table of the alluvial aquifer (poloMčić et al., 2013a, Fig. 6). Groundwater recharge of the alluvial aquifer takes place through infiltration of precipitation and river water. The pliocene sands aquifer is mostly confined and receives re- charge from outside of the boundaries of the ore body, on the northern edge of the study area. The ore body aquifer was formed in limonites which are characterized by signifi- cant porosity and water content. Besides the alluvial aquifer, this is the most significant aquifer in the study area in terms of dewatering. The hydraulic link between the ore body aq- uifer and the river Gomjenica was shown by poloMčić et al., (2013a, Fig. 7). Groundwater in the limonite ore body is indirectly recharged by groundwater from the alluvial de- posits, where they come into contact, and also from the karst aquifer at the bottom of the ore body. The hydraulic link be- tween the ore body aquifer and the alluvial aquifer is shown in Figure 3. The correlation coefficients between groundwa- ter levels of alluvial aquifer and ore body aquifer indicate a very strong link (r2=0.95). The karst aquifer is a confined aquifer (poloMčić et al., 2013a) and formed in Carbon- iferous limestones and dolomites. it is not significantly wa- ter bearing and discharges directly into the limonite ore body in places where these two aquifers are in direct hydraulic contact (BAJić & poloMčić, 2008). Based on the above, it can be concluded that there is a hydraulic link between all the aquifers. Additionally, the ob- jective of chemical analyses (BAJić & poloMčić, 2008) was to provide evidence of such a link between all aquifers in the Buvač mine area. These analyses also indicated that the groundwater in the study area belongs to the hydrocar- bonate class, Ca type. The aquitards are dominated by Quaternary sandy clays, pliocene clays and carboniferous altered siltstone, sand- stone and compact calcite. Sandy clays make up the northern periphery of the omarska-prijedor field, above an altitude of 160 m. They constitute the first and second palaeoterraces of the Gomjenica river and overlie pliocene strata. The pliocene clays along the southern and northern edges of the field are extensive and lie directly on top of palaeozoic sed- iments. The thickness of these clays range from 10 - 60 m. pliocene clays also constitute a large proportion of the min- ing field, apart from the dominant carboniferous altered silt- stone. These impermeable layers are lateral barriers to the groundwater in limestones. The floor of the ore body is also made up of impermeable rocks – carboniferous altered silt- stone, sandstone and compact calcite (Figure 2a). The beginning of ore exploitation in the Buvač mine in 2008 had permanently and irreversibly altered the original hydrogeological and hydrological conditions in the study area. Between 2008–2012 the dewatering system for protec- tion of the open cast mine from groundwater inflow consisted of 8 wells in the alluvial aquifer and 7 wells in the ore body aquifer. There is also a significant number of piezometers – more than 20 which provide data on the groundwater heads. In 2009, the river beds of the River Gomjenica and River Bistrica were diverted into a new river bed with a length of 2,919 m, a width of 20 m with a bank inclination of 1:2. The Figure 3: Correlation dependence between groundwater levels of alluvial aquifer (a) and ore body aquifer (b) Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 127 river bed elevations range from 156.3 to 153.63 m.a.s.l., and the general river bed slope is i=0.0011 %. The diverted river bed is designed for 100 year maximum water levels, and a protective levee was constructed on the left bank which is 1 m higher than the 100 year water level. The new river bed and the natural river beds of the River Gomjenica and River Bistrica are shown in Figure 1. 3. NUMERICAL MODEL (METHODOLOGY) The conceptual model is based on the geological informa- tion obtained from boreholes (Figure 1) and water-level data from observation wells. The model schematization includes a surface clay layer and a basal argillic siltstone of low per- meability and an interbedded sequence of aquitards and four aquifers (poloMčić et al. 2013a). The system is modeled using eight layers. The model layer characteristics are given in Table 1. except for the overlying clays and alluvial grav- els and sands, there is no lithological unit with continuous distribution in one model layer. Table 1 shows lithological units which are present in each of the model layers. The hy- draulic characteristics shown were derived from numerous research works and used in designing the first model (poloMčić et al. 2013a). The three-dimensional finite-difference numerical model for this study was developed using Modflow (HArBAuGH, 2000) with Groundwater vistas as a graphical user interface (ruMBAuGH & ruMBAuGH, 2011). The model domain is 2260 × 2040, and 175 m in depth. The domain is oriented north-south and discretized into 102 rows, 113 columns, and 8 layers, with 85.425 active cells. The cell size of the flow field discretization in plan is 20 m × 20 m this was deemed a reasonable size given the number and quality of available data. The hydraulic characteristics of the model layers were assigned based on the hydraulic conductivity and specific storage values (Table 1) that were previously determined during well testing in the open cast mine (poloMčić et al., 2013a). The previous numerical model of the Buvač open cast mine was updated to reflect the current condition of the min- ing works and the diversion of the river Gomjenica, and supplemented with new data of the groundwater regime from october 2010 – october 2012. 3.1. Boundary and Initial Conditions A head-dependent flux boundary condition (cauchy’s or mixed boundary condition) and a boundary of prescribed flux (Neumann boundary condition) were used in the model. The effect of the diverted course of the river was simulated with the head-dependent flux boundary. in the ModFlow code, this simulation was performed using the river bound- ary condition and it was assigned in the first and second model layer (Figure 4a) with water level values for the pe- riod from 1 october 2010 to 15 october 2012 (Figure 5a). All geometric characteristics of the new river bed of the River Gomjenica were incorporated into this type of bound- ary condition. The initial thickness of the sediment on the river bed bottom was assigned a value of 0.1 m, with an ini- tial value of the hydraulic coefficient of the river sediment Table 1: Model layers and initial values of hydraulic parameters. Model layer Lithostratigraphic unit Hydraulic conductivity (at x,y-axis) (m/s) Hydraulic conductivity (at z-axis) (m/s) Specific storage (1/m) Specific yield 1 Clay overburden 8.50x10–7–1.00x10–6 1.00x10–6 0.001 0.037 2 Alluvial gravel and sandy gravel 1.40x10–4–3.80x10–4 1.00x10–4 2.25x10–5 0.225 3 Clay and sandy clay 6.20x10–6–2.00x10–5 1.00x10–6 5.00x10–5 0.06 Part of ore body 2.30x10–4 – 6.00x10–5 0.06 4 Clay and sandy clay 5.00x10–6 1.00x10–6 5.00x10–5 0.06 Sand and gravel 1.00x10–4–1.40x10–4 1.00x10–4 2.25x10–5 0.23 Part of ore body 2.3x10–4 1.00x10–4 6.00x10–5 0.06 5 Clay and sandy clay 4.0x10–6 1.00x10–6 5.00x10–5 0.06 Part of ore body 2.3x10–4 1.00x10–4 6.00x10–5 0.06 6 Argillic siltstone and sandstone 1.00x10–6 1.00x10–6 6.30x10–5 0.06 Part of ore body 2.3x10–4 1.00x10–4 6.00x10–5 0.06 7 Limonite and fine limonite 5.0x10–5–2.40x10–4 1.00x10–6 5.00x10–5 0.06 Argillic siltstone and sandstone 1.5x10–5 1.00x10–6 6.30x10–5 0.06 Limonite ore body 2.3x10–4 1.00x10–4 6.00x10–5 0.06 8 Limestone and dolomitic limestone, siderite and ankerite in argillic siltstone and siltstone 3.52x10–4–4.70x10–4 2.50x10–4 6.30x10–5 0.06 Argillically altered silstone and sandstone and compact calcite 1.00x10–7 1.00x10–7 6.30x10–5 0.06 Geologia Croatica 68/2Geologia Croatica 128 of 1x10-6 m/s. By assigning head-dependent flux boundary conditions, the model simulates groundwater inflow and out- flow into the model. in the Modflow code, this was repre- sented with a general head boundary (GHB). In the alluvial aquifer to the east and west, this boundary condition simu- lates the outflow or inflow into the model which also gener- ally follows the River Gomjenica outside of the study area (Figures 4 and 6). The ore body during the development of the mine had the role of a local base level of erosion towards which the surrounding groundwater gravitates. Based on ear- lier research, and as a result of the first model (poloMčić et al., 2013a), the existence of impermeable rocks in the south was determined, and the absence of significant inflow into the study area from the north. In the seventh model layer with the largest distribution of the ore body in the east and west, a general head boundary was assigned (Figure 4b and Figure 6). According to the layers shown in Figure 2, in the karst aquifer (7) and the pliocene aquifer (4), the inflow into the model was assigned only in the northern contour of the study area. Also, the GHB hydraulic conditions were taken from the first model (Figure 6). The pliocene aquifer is much larger north of the model domain. The effect of that aquifer on the edges of the pliocene sands included in the model was specified via the GHB boundary condition, simulating inflow while keeping in mind the hypsometric positions of the re- charge zone of this aquifer along the southern slopes of Mt. kozara (north of the model domain). The karst aquifer was handled in a similar manner in the 8th model layer (Table 1), which is represented in the central part of the model domain, while in the northernmost part it extends beyond the model domain. it was also assumed that it was part of a larger aqui- fer to the west, as it was detected in open cast mines west of study area. An additional problem was a lack of data as there are no piezometers in the pliocene and karst aquifers. Figure 4: Distribution map of piezometric levels (15 Oct. 2012) in the alluvial aquifer (a) and levels in the ore body (b) and the boundary conditions in the open cast mine “Buvač”. Figure 5: An overview of groundwater levels: a) in the alluvial aquifer and b) in the limonite ore body in the period 1 October 2010 – 15 October 2012. Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 129 Figure 6: Column (south-north) and row (west-east) cross section. The outer contours of the impermeable units (Layers 3 – 8) are represented by a no-flow boundary condition (Fig- ure 6). initial heads for the karst aquifer and the pliocene aquifer, as there is no exact data, were based on the results of calibration of the first model (poloMčić et al., 2013a). unpublished data were used for the initial heads of these aq- uifers in the first model, which were collected during the course of designing urban and rural water supply systems, at 10 km and 18 km northwest of the study area, respectively. Based on the regional extent of the pliocene and karst aqui- fers, their hypsometric positions and the activities relating to the nearby water supply projects, the conclusion was that these aquifers are slightly pressurized. A total of 8 drainage wells were in operation in the Buvač open cast mine from 1 october 2010 – 15 october 2012. The individual capacities of the wells ranged from 1-50 L/s, with a total capacity of 158 L/s. The locations of the wells in the area included in the model are shown in Figure 4. The wells only drain the alluvial aquifer and the ore body aquifer. The wells were assigned a “prescribed flux” bound- ary condition. with this type of boundary condition, the case of no flow is also assigned (q=0), i.e. selected cells can be designated “no-flow” cells in the Modflow code. These no flow cells represented zones where material has been exca- vated as a consequence of mining (Figures 4). in nature, these zones are practically impermeable. The exception is a part of the excavated ore body the active surface of which is relatively small and continually changes with the dynamics of ore exploitation. Along the working benches of the mine there are drainage canals that direct precipitation to a water collector and from there transfer the water by pumps into the river Gomjenica, so the working area for ore exploitation is not at risk from run off. recharge is simulated by selecting the prescribed flux boundary condition in the first model layer only. This creates an influx as the difference between the total recharge and evapotranspiration. According to the 2007 model (poloMčić et al., 2013a), the average infiltration was 15% of total precipitation; this value was used as an initial condi- tion applied to precipitation for the period 1 october 2010 – 15 october 2012. The existing cut-off wall on the south and east side of the ore body (Figure 4), and a new one, which was modeled in a number of the predictive option models as discussed later, is assumed to be composed of a cement-bentonite fill with a thickness of 1 m and hydraulic conductivity of 1 x 10-8 m/s. This feature was simulated with a prescribed flux boundary condition (Neumann boundary condition), i.e. “wall” (flow barrier) in the Modflow code. The proposed cut-off wall (Figure 4), will extend through the entire alluvium layer down to 1 metre in clayey sedi- ments, its geometric characteristics, as well as the character- istics of the infilling were assigned in the first and second model layer. 3.2. Model recalibration The model was recalibrated under transient flow conditions, with a time step of one day (a total of 654 steps) for the pe- riod 1 october 2010 – 15 october 2012, which in the lower level of iteration was divided into 10 parts, of unequal dura- tion (factor 1.2). Measured piezometric levels in the alluvial aquifer (Figure 5a) and in the limonite ore body aquifer (Fig- ure 5b) were used as calibration targets. Figure 4 shows the plan location of the measuring points and the distribution of the piezometric levels in the alluvial aquifer and in the ore body as of 15 october 2012. Gradually, over time, a certain Geologia Croatica 68/2Geologia Croatica 130 number of piezometers and wells (Figure 5) were unusable due to the lowering of groundwater levels below their max- imum depth. As a confirmation of the quality of the per- formed recalibration of the model, Figure 7 shows the time series and calibration graph. Results of the mathematical simulation of the groundwater regime, in relation to the reg- istered groundwater levels, are in good agreement. The sta- tistical indicators of the quality of model recalibration are: Residual Mean: -0.03, Residual Standard Deviation: 0.09, Absolute residual Mean: 0.08; residual Sum of Squares: 0.15; rMS error: 0.10; Normalized rMS 5.84 %; Minimum residual: -0.22; Maximum residual: 0.12. Under the operating conditions of the dewatering wells system in the mine, a change in the flow conditions occurred i.e. intensified three-dimensional groundwater flow as a result of the diversion of the River Gomjenica into a new river bed and the development of the mine. Table 2 gives an overview of the basic elements of the groundwater water balance after model recalibration. in the alluvial aquifer, the predominant mode of recharge is water infiltration from the river Go- mjenica (78.17%). in relation to the first model (poloMčić et al., 2013a) which was calibrated to conditions prevailing in 1973, river water recharge to the alluvial aquifer has increased due to the operation of drainage wells which have a total dis- charge of 158 l/s. The flow between the existing aquifers in the mine has also changed since 1973. Model predicted in- flows to the ore body from the alluvial aquifer and from the karst aquifer are 26.8 l/s and 8.3 l/s respectively. inflow to the ore body is greatest from the east, with less from the west. The total inflow into the ore body decreased by 40.1 l/s in relation to the quantities of water captured by wells, which confirms the successful drainage of the ore body up to this point (Figure 5b). The balance of the pliocene aquifer is not significant for the total groundwater balance in the study area. inflow from the karst aquifer to the ore body aquifer is par- tially replenished by groundwater inflow from the north. 3.3 Management Scenarios The recalibrated groundwater model was used to evaluate options for water management for 2013–2024. Calculations were based on sequential time steps; the result at the end of Table 2: Groundwater balance (15 October 2012). Aquifer Boundary condition Inflow to model (L/s) Outflow from model (L/s) Alluvial aquifer and aquifer in pliocene sands Effective infiltration 15.2 Gomjenica river 97.4 GHB – east 2.3 GHB – west 9.7 Drainage wells 22 Migration into the ore body aquifer 26.8 The limonite ore body aquifer GHB – east 36.7 GHB – west 24.1 Drainage wells 136 Karst aquifer GHB – east 8.3 GHB – north 7.9 Sum 193.3 193.1 Figure 7: Time series and calibration graph for the analyzed period (1 October 2010 – 15 October 2012): a) alluvial aquifer b) ore body aquifer. Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 131 one time interval representing the initial condition for cal- culation in the next. The calculations were performed in a transient flow regime, using a month as the basic time inter- val for the calculation within one year. At the end of each year, in prognostic calculations, there has been increase in area of “no flow” cells as a result of the yearly advancement of the mining works – excavating the spoil and extracting the ore, and the development of analyzed dewatering sys- tems. Three groundwater management scenarios were mod- eled, as described later, and Figure 8 shows the elements of the dewatering system in the „Buvač“ open cast mine: on the day of recalibration of the hydrodynamic model (a), the de- watering system according to variant 1 (b), variant 2 (c), and variant 3 (d). All three scenarios were identical during 2013 and 2014, while for 2015 scenarios 2 and 3 are identical. From 2016, all scenarios are different and develop indepen- dently. For all three scenarios, the open cast development follows the planned mining activities until the end of 2024. All three management scenarios included the following features: • 8 existing drainage wells in the west that drain the al- luvial aquifer, • the existing drainage channel Du-4 in the alluvial aq- uifer; • starting from 01.01.2015 the Gomjenica river is partly relocated to the new riverbed north of the mine; • the river Gomjenica diversion; • the locations, initial capacities, and length of operation of drainage wells that primarily drain the ore body, and • the main mining criterion is that the groundwater level should be at least 15 metres below the working bench in the ore body. The drainage channel Du-4 is assumed to be excavated to the bottom of the alluvial sands and gravels in the north- ern part of the mine, ahead of the mining. The channel will intercept groundwater which is not extracted by the wells. The water from the drainage channel flows to the water col- lector and is then returned to the River Gomjenica. At the beginning of prognostic calculations, seven drain- age wells are assumed in the ore body each with a capacity of 8-50 l/s, for a total capacity of 126 l/s. Scenario 1 According to Scenario 1, 33 additional wells in the al- luvial aquifer to the north of the open cast will start on 1 January 2015, will increase the protection from river water seepage based on the 8 exisiting wells. The initial well ca- pacities range from 2 L/s to 7 L/s (with a sum of 107 L/s) and are assumed to operate continually during the entire life of the mine, i.e. until the end of 2024. Figure 8b shows the dewatering system according to variant 1. Scenario 2 In Scenario 2, instead of the 33 drainage wells, a dewa- tering system in the alluvial aquifer comprises a cut-off wall, almost 2000 m long and extending 1m into the clayey silty sediments below the alluvial sediments. The cut-off wall is assumed to be completed on 1 January 2016. Figure 8c shows the dewatering system according to variant 2. Scenario 3 Scenario 3 represents a combination of the previous two scenarios. In the northwest part of the mine, where the open cast is closest to the river, a 1000 m long cut-off wall is as- sumed, similar to that described in scenario 2. The cut off starts near the existing well eb-8, and is assumed to be com- plete on 1 January 2016. continuing from the wall, 13 drain- age wells are assumed to be constructed with the same char- acteristics and initial capacities as in scenario 1, for a total capacity of 65 l/s. The wells begin operation on 1 January 2017, and operate until the end of mining (2024). Figure 8d illustrate this scenario. 4. RESULTS AND DISCUSSION Before analyzing the results of the simulation of different variants of the dewatering system, it should be noted that in Bosnia and Herzegovina, it is usual for projected mining works to be presented on a yearly state (31 December) dur- ing the first five years, and subsequently on a 5-year level with a projected final contour of the mine at the end of ore exploitation. The most significant change in the flow pattern in the wider area of the mine depends on the construction dynamics of the advancing excavation contour, given in 1-year time steps (at the end of the calendar year, 31 Decem- ber). It is realistic for the mine advancement to be a linear continuous process. This applies especially to the overview of derived results per time interval viewed in 2022 (a five- year step of 2017-2022) and 2024 (a two-year step: 2022- 2024). it is also necessary to provide an explanation, regard- ing the correct interpretation of the calculated results, which relates to the operation of drainage wells. The wells were as- signed in the model using their discharge, for which the ini- tial values represent one of the results of the prognostic cal- culations. The results of the calculations proved to be very sensitive to changes in initial well capacities, which must gradually decrease over time. otherwise, there is an exces- sive decrease in the piezometric head in the wells. if too large initial capacity were given, this would result in shut- ting down of the well. These facts primarily apply to the hy- draulic characteristics of porous media. In the other hand, these are the consequences of the limits of the MoDFlow finite difference model. First is, that it is a quasy 3D flow model. Second, the hydraulic conductivity tensor is repre- sented only with three principal components kxx, kyy and kzz perpendicular to the faces of the finite difference cells (horizontally rectangular). All of this is not representing in the best possible way 3D problems which occur in the vicin- ity of the pumping wells. The piezometric heads in the allu- vial aquifer obtained for each scenario are shown in figures 9 (a, b) and 10 (a, b, c) and the piezometric heads in the ore body are shown in figures 9c and 10d, at the end of 2015 and 2024. over time, the operation of the drainage system, ex- cavation of overburden and of the ore body, will lead to a Geologia Croatica 68/2Geologia Croatica 132 gradual lowering of the regional piezometric heads. During one years simulated operation of the drainage system, be- sides the “no flow” cells, (which are primarily a consequence of mining), “dry cells” appear in the model and expand in time due to the operation of the system. The protective influence of the system on groundwater ingress to the open cast mine is manifested by different gra- dients of hydraulic head both within and outwith the outer contours of the drainage system. In relation to 2013 and 2014, when the system in the alluvial aquifer consists only of the DU-4 drainage channel (Figures 9 and 10), there be- gins in 2015 or 2016, a significant change in the flow pattern for each of the three simulated dewatering systems. According to the first simulated scenario (Figures 9a and 10a) for protection of the open cast mine, a total of 41 fully penetrating wells together with with the drainage channel Du-4, compose the drainage system. The first scenario is the least effective, and results in the smallest decrease in ground- water levels ahead of mining. The second scenario (Figures 9b and 10b) with a cut-off wall which prevents groundwater inflow through the alluvial sediments from the direction of the river Gomjenica, very quickly results in dewatering of the alluvial deposits ahead of mine development, and causes the Du-4 drainage channel to cease surface flow relatively quickly. over time, the “dried” surface spreads, while in the north-east of the mine a narrow zone remains along which there is inflow into the open cast mine, as the result of the flow of uncaptured groundwater between wells eb-1 to eb- 8. According to the achieved protective effects, this variant represents the most efficient defense from the inflow of groundwater into the open cast mine through the alluvial de- posits. The third scenario (Figures 9b and 10c) simulates the operation of the open cast mine protection system which is composed of the existing drainage wells eb-1 to eb-8 in the west, a cut-off wall in front of the open cast mine where the river bed of the River Gomjenica is closest, the drainage wells eb-29 to eb-41 continuing after the wall in the north, as well as the DU-4 drainage channel. The third scenario is a combination of the first and second scenarios and the im- pact on groundwater flow is intermediate between scenarios 1 and 2. Figure 8: Elements of the dewatering system in open pit mine „Buvač“: on the day of re-calibration of the hydrodynamic model (a), the dewatering sys- tem according to variant 1 (b), variant 2 (c), and variant 3 (d). Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 133 In terms of drainage of the ore body, an almost identical decrease in groundwater levels is predicted for all three of the analyzed scenarios (Figures 9c and 10d), due to the iden- tical drainage system in the ore body for each scenario. The obtained differences per scenario are only related to the ca- pacities of the individual wells (Figure 11). The decrease of groundwater levels in the ore body is, for all variants and for each analyzed year, below the working bench levels in the assigned values (a minimum of 15 metres below). The effects of the three scenarios are analyzed for the alluvial aquifer and for the ore body (Figure 11) through comparisons of the predicted flow of the drainage wells. All three analyzed scenarios include a line of wells (eb-1 do eb- 8) in the west. in the first scenario, these wells have greater predicted flow than in scenarios 2 and 3, due to the absence of the cut-off wall. The maximum well capacity will be reached at the beginning of 2015 (158 L/s) and towards the end of the simulated period will decrease by 14%. In sce- narios 2 and 3, the greatest differences in well flow occur in those wells closest to the cut-off wall, due to the restricted inflow of water to the wells. in scenarios 2 and 3, the pre- dicted flows from wells in the west are 40% less at the end of the calculation period, in comparison to scenario 1. There is a similar occurrence in other wells close to the cut-off wall in scenario 3 (wells eb-29 and eb-30). in wells eb-38 to eb- 41 in variant 3, the proximity of the cut-off wall does not af- fect a significant decrease in the capacity of these wells, due to the proximity of the diverted river bed of the river Go- mjenica, and the small initial capacities of the wells. Drain- age channel Du-4 is likely to be most effective in the first operational years of the model with an inflow of 38 l/s, where, apart from the drainage channel and drainage wells, there are no other facilities to protect the mine from ground- water inflow through the alluvium. The inflow into the chan- nel varies between scenarios with the smallest flow predicted for scenario 2 (no inflow until January 2016) and the largest flow for scenario 1 (16 l/s in January 2016 to 5 l/s in Sep- tember 2017). Figure 9: The distribution of piezometric levels: a) in the alluvial aqui- fer at the end of 2015 (Scenario 1), b) in the alluvial aquifer at the end of 2015 (Scenario 2 and 3) and c) in the ore body at the end of 2015. Geologia Croatica 68/2Geologia Croatica 134 The differences between the scenarios in terms of water balance in the ore body are practically negligible and a sin- gle conclusion may be drawn for all three. A number of “Bu wells”, even after a periodic reduction, are withdrawn from operation, due to the overall decrease in groundwater levels (Figure 11). From the beginning of 2013 to the end of 2015, the total capacity of the wells in the ore body (BU wells) de- creased by 13% (from 121 l/s to 105 l/s). A certain increase in the predicted flow (Figure 11) of the wells in the ore body is due to the introduction of new wells (January 2016). From 2016 when the dewatering system in the ore body reaches the maximum capacity of 124 l/s, to the end of the prognos- tic period, there is a continual decline in well capacity as well as in mining activities, and the deactivation of a number of wells. At the end of 2024, the total well capacity is de- creased by 68%. in the first three years the model predicts similar ground- water discharge into the pit of about 10–14 L/s with all sce- narios showing similar inflow. The inflow reduces over the modeled period to about 6 - 10 l/s for scenario 1 and 3 - 6 l/s for scenario 2 with scenario 3 inflows being intermedi- ate between the scenario 1 and scenario 2 inflows. 5. CONCLUSION The exploitation of the limonite ore in the open cast mine Buvač began at the end of 2008. New hydrogeological re- search and monitoring of the groundwater regime were per- formed to assist in the selection of groundwater management schemes for protection of the Buvač open cast mine from groundwater inflow until the end of ore exploitation (2024). The numerical model provides a tool for predictive calcula- Figure 10: The distribution of piezometric levels: a) in the alluvial aquifer at the end of 2024 (Scenario 1), b) in the alluvial aquifer at the end of 2024 (Sce- nario 2), c) in the alluvial aquifer at the end of 2024 (Scenario 3) and d) in the ore body at the end of 2024. Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 135 tions to compare options for protection of the open cast mine from groundwater inflow. Three groundwater management scenarios were modeled for the period from 2013 to 2024. The scenarios mainly differ in the methods of management of groundwater flow in the alluvial aquifer. in all of the ana- lyzed scenarios of the dewatering system, the required drop of groundwater levels to more than 15 metres below the working bench in the ore body, has been achieved for every year in the prognostic calculations,. All the facilities protecting the open cast mine from wa- ter ingress via the diverted river bed of the River Gomjenica are positioned parallel to the river course, north of the mine. Scenario 1 models management of the alluvial flows by the use of a line of wells, while scenarios 2 and 3 include differ- ent numbers of wells and variable lengths of cut-off wall. The design and sequence of construction of the drainage wells in the ore body is the same for all options. All options show that groundwater inflow to the pit, and flow through the alluvial aquifer decreases as a consequence of the dewa- tering systems. The projected open cast mine contours are the primary basis for the analysis of the different dewatering systems. in Bosnia and Herzegovina, and in some of the former yu- goslav states, it is usual to present projected mining works on a yearly state (year end) for the first five-year period, and then on a 5-year interval level, with a projected final open cast mine contour at the end of ore exploitation. like- wise, after each five-year period of ore exploitation, the prognoses of the dewatering system operation are verified or corrected. The most significant change in the flow pat- tern in the wider area of the mine is influenced by the con- struction dynamics of the excavation contour advancement, which is presented with a time step of a minimum of one year, which differs from the real conditions when advance- ment of the mine is linear in process. As a reference posi- tion, the prognostic calculations adopted the condition of the mine contour at the end of the year (31 December) which is kept during the entire following year of calcula- tions. More significant differences appear in 2022 (after five years) and in 2024 (after two years). Besides this, in all the analyzed dewatering systems, the drainage wells have a significant role. During the calculations, the results proved to be very sensitive to changes in the initial values of well capacities, which must decrease gradually over time. otherwise, there is an excessive decrease in piezo- metric heads in the wells which results in wells either be- ing shut down or showing an insufficient level of reduction in the wider area of the wells. Therefore an especially sen- sitive task during prognostic calculations was assigning the discharge of the drainage wells, where several criteria and limits had to be kept in mind: • The initial well capacities should not be too large, as the dewatering effect which is aimed at will not be achieved – even with a steep decrease in the ground- water levels in a well, a corresponding drawdown does not form in the surrounding area. This is a not only a result of the hydraulic characteristics of the media (al- luvial sediments and the limonite ore body), but also their variable thickness. • The life span of the wells which are located in the course of the mine advancement front, depends on the dynamics of the advancement of spoil excavation. over time, with the advancement of the spoil benches, the well gradually shortens and its capacity decreases, until the final shut down, i.e. destruction of the well. • The well discharge decreases correspondingly and si- multaneously with the gradual drainage and decreasing groundwater levels. It is necessary to create a balance between the well discharge and level reduction, in the well itself and its immediate surroundings, as well as in the wider zone of the well. Figure 11: A comparative diagram of the predicted total flow of drainage wells and in the DU-4 channel (January 2013 – December 2024) for all scenarios where EB wells are in the alluvial aquifer and BU wells are in the ore body. Geologia Croatica 68/2Geologia Croatica 136 • A proportion of the wells in the ore body are not de- stroyed, remaining permanently in operation, and their capacities decrease over time. • A number of wells are removed from operation over time due to the decrease in groundwater levels below the elevation of the well screen construction. From a hydrodynamic aspect, Scenario 2 which includes a cut-off wall ahead of mining of the north wall of the open cast mine provides the best protection from the inflow of groundwater from the alluvial aquifer from the direction of the diverted river bed of the River Gomjenica. Scenario 3 is a combination of the first and second scenarios, with a de- watering system composed of a shorter cut-off wall and drainage wells, and this is the second most effective option. A techno-economic analysis is required to select the best of the analyzed scenarios for protection of the open cast mine “Buvač” from groundwater inflow. Finally, it is concluded that the application of ground- water modeling has proven very useful in comparing meth- ods for protection of the open cast mine from groundwater inflow. The hydrodynamic model of the aquifer regime and management scenarios for protection from groundwater in- flow in the open cast mine “Buvač” opens the way toward system optimization and sustainable economic development. ACKNOWLEDGMENT our gratitude goes to the Ministry of education, Science and Technological Development of the Republic of Serbia for financing projects “oi176022“, „Tr33039” and „iii43004“. Two reviews, by kristijan poSAvec (zagreb, croatia) and an anonymous reviewer, are also acknowledged for their contribution and effort in providing constructive criticism and improving the manuscript. REFERENCES BAJić, D. & poloMčić, D. (2008): 3D Hydrogeological model of the limonite ore body “Buvač” (republika Srpska, Bosnia and Herze- govina). in proceedings of the iv international conference “coal 2008” held at Belgrade, Serbia, october 2008) – yugoslav opencast Mining Committee, 2008. BAHrAMi, S., ArDeJANi, F.D., ASlANi, S. & BAAFi, e. (2014): Numerical modelling of the groundwater inflow to an advancing open pit mine: kolahdarvazeh pit, central iran.– environmental Monitoring and Assessment, doi: 10.1007/s10661-014-4025-x BANkS, D., Frolik, A., Gzyl, G. & roGoż, M. (2010): Modeling and monitoring of mine water rebound in an abandoned coal mine complex: Siersza Mine, upper Silesian coal Basin, poland.– Hydro- geology Journal, 18/2, 519–534. doi: 10.1007/s10040-009-0534-z BoSkiDiSA, i., piSiNArASA, v., peTAlASA, c. & TSiHriNTzi- SA, v.A. (2012): Monitoring and modeling of two alluvial aquifers in lower Nestos river basin, Northern Greece.– Journal of environ- mental Science and Health, part A: Toxic/Hazardous Substances and environmental engineering, 47/12, 1849–1868. doi: 10.1080/ 10934529.2012.689552 BrAwNer, c.o. (1982): Control of groundwater in surface mining.– Mine water and the environment, 1/1, 1–16. doi: 10.1007/ BF02504603 DoHerTy, e.J. & HuNT, J.r. (2010): Approaches to Highly parame- terized inversion: A Guide to using peST for Groundwater-Model calibration.– u.S. Geological Survey Scientific investigations re- port 2010-5169: reston, vA, uSA, 59 pp. GöDeke, S. (2011): Simulation of groundwater mounding at a west Australian mine site.– environmental earth Science, 64/5, 1363– 1373. doi: 10.1007/s12665-011-0961-8 HANNA, T.M., AzrAG, e.A. & ATkiNSoN, l.c. (1994): Use of an analytical solution for preliminary estimates of groundwater inflow to a pit.– Mining engineering, 46/2, 149–152. HArBAuGH, A.w., BANTA, e.r., Hill, M.c. & McDoNAlD, M.G. (2000): MoDFlow-2000, the u.S. Geological Survey Mod- ular Ground-water Model: user guide to modularization concepts and the ground-water flow process.– u.S. Geological Survey open- File report 00-92: reston, vA, uSA, 121 p. JiANG, S., koNG, X., ye, H. & zHou, N. (2013): Groundwater de- watering optimization in the Shengli no. 1 open-pit coalmine, inner Mongolia, china.– environmental earth Science, 69, 187–196. doi: 10.1007/s12665-012-1946-y li, T., SoNG, H., HuANG, G., Bi, y. & li, H. (2014): Assessment of groundwater changing trends through the generalized large well method with confined-unconfined flow model in open-pit mine area.– environmental earth Science, doi: 10.1007/s12665-014-3357-8 MArANDi, A., kArro, e., polikArpuS, M., JõeleHT, A., ko- Hv, M., HANG, T. & HiieMAA, H. (2013): Simulation of the hy- drogeologic effects of oil-shale mining on the neighbouring wetland water balance: case study in north-eastern estonia.– Hydrogeology Journal, 21/7, 1581–1591. doi: 10.1007/s10040-013-1032-x MAriNelli, F. & Niccoli, w.l. (2000): Simple analytical equations for estimating ground water inflow to a mine pit.– Ground water, 38/2, 311–314. doi: 10.1111/j.1745-6584.2000.tb00342.x MylopouloSA, N., MylopouloSB, y., kolokyTHAB, e. & TolikAS, D. (2007): Integrated water management plans for the restoration of lake koronia, Greece.– water international, 32/1, 720–738. doi: 10.1080/02508060.2007.9671993 poloMčić, D., BAJić, D., pApić, p. & SToJković, J. (2013a): Hy- drodynamic model of the open-pit mine “Buvač” (republic of Srp- ska).– Journal of Sustainable Development of energy, water and environment Systems 1/3, 260–271. doi: 10.13044/j.sdewes.2013. 01.0019 poloMčić, D., HAJDiN, B., STevANović, z., BAJić, D. & HAJ- DiN, k. (2013b): Groundwater management by riverbank filtration and an infiltration channel: The case of obrenovac, Serbia.– Hy- drogeology Journal, 21/7, 1519–1530. doi: 10.1007/s10040-013- 1025-9 QiAo, X., li, G., li, M., zHou, J., Du, J., Du, c. & SuN, z. (2011): influence of coal mining on regional karst groundwater system: a case study in west Mountain area of Taiyuan city, northern china.– environmental earth Science, 64, 1525–1535. doi: 10.1007/s12665- 010-0586-3 ruMBAuGH, J.o. & ruMBAuGH, D.B. (2011): Guide to using Groundwater vistas: version 6. environmental Simulations, New york. rApANTovA, N., GrMelA, A., voJTek, A., HAlir, J. & MicHA- lek, B. (2007): Ground water Flow Modelling Applications in Mining Hydrogeology.– Mine water and the environment, 26/4, 264–270. doi: 10.1007/s10230-007-0017-1 SHeveNell, l. (2000): Analytical method for predicting filling rates of mining pit lakes: example from the Getchell Mine, Nevada.– Mining engineering, 52/3, 53–60. STruziNA, M., Mueller, M., DreBeNSTeDT, c., MANSel, H. & JolAS, p. (2011): Dewatering of Multi-aquifer unconsolidated Polomčić and Bajić: Application of Groundwater modeling for designing a dewatering system: Case study of the Buvač Open Cast Mine, Bosnia and Herzegovina Geologia Croatica 137 rock opencast Mines: Alternative Solutions with Horizontal wells.– Mine water and the environment, 30, 90–104. doi: 10.1007/ s10230-011-0149-1 STrzoDkAA, k. & FiScHerA, M. (1988): Dewatering wells and sealing walls for water control in lignite mines.– International Jour- nal of Surface Mining, reclamation and environment, 2/2, 105– 107, doi 10.1080/09208118808944143 ŠuBArANović, T., pAvlović, v., poloMčić, D. (2013): influ- ence on environment of Sealing Screen at lignite opencast Mines.– Metalurgia international, 18/5, 169–176. wArNer, J.w., kHAzAeiB, w., wArNerc, J., MANGHiD, F., pHrANere, r.w., MorTAzAvie, B. & NAMvArF, r. (2009): Flow and transport modelling of a highly stressed aquifer to refine management strategies.– water international, 34/2, 264–279. doi: 10.1080/02508060902894622 wu, Q., DoNG, D.l., SHi, z.H., wu, X., SuN, w.D., ye, G.J., li, S.w. & liu, J.T. (2000): optimum combination of water drainage, water supply and eco-environment protection in coal-accumulated basin of North china.– Science in china Series D: earth Sciences, 43/2, 122–133, doi 10.1007/BF02878141 wu, Q., Hu, B.X., wAN, l. & zHeNG, c. (2010): Coal mine water management: optimization models and field application in North china. Hydrological Sciences Journal, 55/4, 609–623. doi: 10.1080/02626661003798310 Manuscript received December 08, 2014 Revised manuscript accepted April 01, 2015 Available online June 16, 2015